Enterprise software deployments have long moved at a predictable—and painfully slow—pace: demos, security reviews, proof-of-concepts, custom data pipelines, then months before production. For master data management (MDM), which unifies fragmented records into a trusted source of truth, the process has been especially complex.
Databricks aims to compress that timeline with Apps on Marketplace, now in public preview. The feature lets customers discover, review, and install third-party apps directly within their Databricks workspace, where they run on a secure serverless runtime. LakeFusion, an MDM provider, is among the 20 launch partners.
The Install is the Easy Part
Some Marketplace listings, including certain apps, are by request and require provider approval before installation can proceed, according to Databricks documentation. Once approved, the customer agrees to terms, configures the requested resources and permissions, and installs the app directly into their environment. Marketplace listings can also include datasets, AI models, notebooks, and MCP servers, per Databricks’ marketplace documentation.
Roz King, Chief Architect at LakeFusion, describes what this replaces: “Deploying a new data management platform has traditionally meant per-customer infrastructure, manual onboarding, and separate security reviews before a single record gets resolved. Through the Databricks Marketplace, customers go from discovery to a running LakeFusion environment in minutes.”
That claim holds up on the infrastructure question specifically. Databricks Apps run on Databricks’ own serverless computer and integrate with platform services already present in the customer’s account, according to Databricks’ developer documentation, which removes the need for a vendor to provision separate infrastructure. The “minutes” framing applies to that installation step once approval is granted. It does not describe the sales and approval cycle before it, which will still vary by organization, or the configuration and data work that follows.
Why Master Data Management Puts the Model to the Test
MDM is a useful category for testing the model because it touches records across multiple enterprise systems. LakeFusion describes legacy implementations as often requiring data to move into new environments, new pipelines to be built, separate governance layers to be stood up, and additional security reviews to clear, before any matching or merging of records can begin, for a task that is conceptually simple: recognizing that “Acme Corp” and “ACME Corporation” in two different systems are the same customer.
LakeFusion MDM works to resolve that kind of duplication across customers, suppliers, providers, products, locations, and assets, producing what the industry calls a golden record: a current, validated single source of truth for an entity, built from the fragmented systems of record that already exist across a business, as IBM describes the concept. If Marketplace-native distribution can shorten the path from duplicate records to a governed golden record, it says something about how much of MDM’s traditional timeline was infrastructure rather than the harder work of governance itself. Whether it does so at scale, and for real production data rather than a test environment, is not yet demonstrated by a public preview.
A Model Still Being Proven, Not a Foregone Conclusion
Approval workflows, permissioning, and provider review remain part of the process by design. What changes is where that work happens: inside a Marketplace flow that a customer’s admin controls, rather than spread across a vendor-specific onboarding project. Stephen Orban, SVP of Product Ecosystem and Partnerships at Databricks, points to demand from customers directly: “Customers consistently ask us for trusted, governed ways to manage and share master data across their organizations as they scale AI.”
That demand describes an appetite for faster, more governed data infrastructure. It does not resolve the harder question of whether a golden record produced through a faster install is one that a business is willing to run production decisions on, which still depends on data quality, stewardship, and testing that no installation flow shortens.
The test that matters is not whether MDM software can be listed in a marketplace and installed through a standardized Marketplace flow. It is whether a governance-heavy system can move from approval to production without recreating the infrastructure burden the marketplace was built to remove. Twenty partners and one public preview are not enough to answer that. They are enough to say the question is now worth asking with specifics, rather than assuming the answer either way.
